results · web viewusing tree species inventories to map biomes and assess their climatic overlaps...
Post on 07-Jul-2020
2 Views
Preview:
TRANSCRIPT
Using tree species inventories to map biomes and assess their climatic overlaps in lowland
tropical South America
Pedro Luiz Silva de Miranda1*, Ary Oliveira-Filho2, R. Toby Pennington3,4, Danilo M. Neves5,
Timothy R. Baker6, Kyle G. Dexter1,3
Short-running tittle: South American Biomes
1School of GeoSciences, University of Edinburgh, Edinburgh, EH9 3FF, UK2Departamento de Botânica, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais, 31270–901, Brazil
3Royal Botanic Garden Edinburgh, 20a Inverleith Row, Edinburgh, EH3 5LR, UK
4Department of Geography, University of Exeter, Rennes Drive, Exeter, EX4 4RJ, UK
5Department of Ecology and Evolutionary Biology, University of Arizona, Tucson 85721, USA
6School of Geography, University of Leeds, Leeds, LS2 9JT, UK
*Correspondence: P.L.Silva-De-Miranda@sms.ed.ac.uk
Phone: +44 7871462345
Address: Room 401, Crew Building, King’s Buildings, Edinburgh EH9 3FF
Key-words: Amazon Forest, Atlantic Forest, Cerrado, Chaco, Cluster Analysis, NeoTropTree,
Savanna, Seasonally Dry Tropical Forest,
Number of words in abstract: 287
Number of words in main text: 5620
Number of references: 84
1
2
3
4
5
6
78
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Abstract
Aim
To define and map the main biomes of lowland tropical South America (LTSA) using data
from tree species inventories and to test the ability of climatic and edaphic variables to
distinguish amongst them.
Location
Lowland Tropical South America (LTSA), including Argentina, Bolivia, Brazil, Ecuador,
Paraguay, Peru and Uruguay.
Time Period
Present
Major Taxa Studied
Trees
Methods
We compiled a database of 4,103 geo-referenced tree species inventories distributed across
LTSA. We used a priori vegetation classifications and cluster analyses of floristic composition
to assign sites to biome. We mapped these biomes geographically and assessed climatic
overlaps amongst them. We implemented classification tree approaches to quantify how
well climatic and edaphic data can assign inventories to biome.
Results
Our analyses distinguish savanna and seasonally dry tropical forest (SDTF) as distinct
biomes, with the Chaco woodlands potentially representing a third dry biome in LTSA.
Amongst the wet forests, we find that the Amazon and Atlantic Forests may represent
different biomes as they are distinct in both climate and species composition. Our results
show an important environmental overlap amongst biomes, with error rates to classify sites into
biomes of 19-21% and 16-18% when only climatic data and with the inclusion of edaphic data,
respectively..
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
Main Conclusions
Tree species composition can be used to determine biome identity at continental scales. We
find high biome heterogeneity at small spatial scales, likely due to variation in edaphic
conditions and disturbance history. This points to the challenges of using climatic and/or
interpolation-based edaphic data or coarse resolution, remotely-sensed imagery to map
tropical biomes. From this perspective, we suggest that using floristic information in biome
delimitation will allow for greater synergy between conservation efforts centred on species
diversity and management efforts centred on ecosystem function.
Key-words: Cluster Analysis, Atlantic Forest, Amazon Forest, Chaco, Savanna, Cerrado,
Seasonally Dry Tropical Forest, NeoTropTree.
Introduction
The biome concept has existed for over a century with the overarching purpose of
delimiting recognisable, ecologically meaningful vegetation units. Humboldt (1816) used the
term phytophysigonomy when referring to areas that may be geographically disjunct, but
share similar vegetation physiognomy or structure. The link between vegetation structure
and climatic conditions was detailed by Schimper (1903), who attributed these similarities to
physiological and anatomical adaptations to precipitation and temperature. The relationship
between vegetation form and climate permeates the majority of vegetation classification
schemes proposed during the 20th century (Clements, 1916; Holdridge, 1947; Walter, 1973;
Whittaker, 1975), and climate is still regarded as the main driver of plant and biome
distributions (Box, 1995; Prentice et al., 1992; Prentice, 1990). More recently, biomes have
been used to categorise the function of ecosystems at large spatial scales, including across
continents (Higgins, Buitenwerf, & Moncrieff, 2016; Woodward, Lomas, & Kelly, 2004), and
the most prevalent biome concept at present, which we employ here, is that of a
widespread vegetation formation with distinct ecosystem function.
The term ‘biome’ itself was first employed by Clements (1916) when referring to the biotic
community, or set of species, occupying a certain habitat. However, subsequently,
Holdridge (1947), Walter (1973), Whittaker (1975) and Odum (1975) gave more emphasis to
the relationship between climate and vegetation structure when proposing classification
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
systems for vegetation formations or biomes, and distanced themselves from the
community composition perspective suggested by Clements (1916). These latter authors
delimited biomes using standard climatic variables, such as mean annual temperature
(MAT) and mean annual precipitation (MAP) (e.g., Whittaker 1975). A motivating factor for
these studies was to create practical classification systems that allow researchers to assign
sites to biome by simply knowing the MAT and MAP (e.g., as in Qian, Jin, & Ricklefs, 2017;
Siepielski et al., 2017). More recently, large-scale remotely sensed data have become
available, which has led researchers to map biomes using simple characterisations of
vegetation physiognomy or ecosystem function, including average vegetation height,
percent tree cover, primary productivity and phenology (Higgins et al., 2016; Hirota,
Holmgren, Van Nes, & Scheffer, 2011; Staver, Archibald, & Levin, 2011; Woodward et al.,
2004). However, remote sensing approaches can fail when biomes are indinstinguiable from
satellite images (Beuchle et al., 2015) or when there is high structural heterogeneity within
biomes (Särkinen, Iganci, Linares-Palomino, Simon, & Prado, 2011)
Meanwhile, the different global biome schemes, be they derived from climate or remote
sensing, often fail to agree on which are the main biomes (e.g Whitakker, 1975 vs. Friedl et
al., 2002 vs. Woodward et al., 2004 vs. Higgins et al., 2016), and can differ dramatically on
the mapping of any given biome (Särkinen et al., 2011). Furthermore, the degree to which
biome maps actually delimit the spatial distribution of ecosystem function is debated
(Moncrieff, Hickler, & Higgins, 2015). The need for more ecologically meaningful definitions
of biomes has led some to suggest that functional traits, such as wood density or leaf mass
per area of the dominant plant species, should be used to define and delimit biomes (Van
Bodegom, Douma, & Verheijen, 2014; Violle, Reich, Pacala, Enquist, & Kattge, 2014). In
order to map functional trait distributions at large spatial scales, researchers have used geo-
referenced collection localities for species with available trait data (e.g. Engemann et al.,
2016; Lamanna et al., 2014). There are challenges with this approach, most importantly, the
absence of trait data for many species, especially in tropical vegetation (Baker et al., 2017;
Sandel et al., 2015; Violle, Borgy, & Choler, 2015). The premise of this paper is that species
occupying distinct biomes have different functional traits and therefore that floristic
information can be used to map biomes, avoiding the uncertainties associated with linking
species composition to trait databases. Species distribution modelling (a.k.a. ecological
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
niche modelling) of indicator species can be used to map biomes (as in Prieto-Torres &
Rojas-Soto, 2016; Särkinen et al., 2011), but such distribution modelling usually uses only
climatic variables as predictors and therefore is subject to similar concerns as mapping
biomes directly based on climatic data. We argue that, at least for some regions, there are
now sufficient species distribution data to map biomes directly using the distribution data
themselves.
The mapping of biomes based on floristic information also offers the possibility of synergies
with conservation (Whittaker et al., 2005). Bioregionalisation schemes that partition space
into geographic units based on species composition and environmental data, such as the
global ecoregions proposed by Olson & Dinerstein (1998) and Olson et al. (2001) – recently
reviewed and updated by Dinerstein et al. (2017) – have been used by researchers and
decision makers in conservation at local and global scales. For example, it was by relying on
Olson & Dinerstein’s (1998) scheme that Myers et al.(2000) and Mittermeier et al. (1998,
2004) proposed the global biodiversity hotspots, which are biomes or geographic subsets of
biomes (i.e. ecoregions), that present high numbers of endemic species and are particularly
threatened.
Brazil, which comprises the majority of the land surface of Lowland Tropical South America
(LTSA), has proposed its own bioregionalisation scheme, the Domain system, established by
Veloso, Rangel Filho, & Lima (1991) and IBGE (2012). The six Domains, which are used to
guide conservation and management policy, are the Amazon Forest, Atlantic Forest,
Cerrado, Caatinga, Pantanal and Pampa. The first two are wet forests, with the Amazon
Forest occupying much of northern LTSA and the Atlantic Forest occurring along the Atlantic
coast of South America, principally in Brazil. They are separated by a ‘Dry Diagonal’ of
seasonally dry forests, woodlands and savanna vegetation formations (Neves, Dexter,
Pennington, Bueno, & Oliveira Filho, 2015; Vanzolini, 1963). The Cerrado Domain is
comprised primarily of savanna and sits in the centre of the Dry Diagonal, occupying much
of central Brazil, but there are disjunct patches of savanna found elsewhere in LTSA,
particularly within the Atlantic and Amazon Forests (Ratter, Ribeiro, & Bridgewater., 1997).
Wet forests intrude into the Cerrado as gallery forests along river courses (Oliveira-Filho &
Ratter, 1995). The Caatinga Domain at the northeast corner of the Dry Diagonal represents
the largest extent of seasonally dry tropical forest (SDTF) in LTSA (Prado & Gibbs, 1993).
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
However, SDTF also occurs in disjunct patches throughout the Cerrado on more fertile soils
(DRYFLOR, 2016; Pennington, Prado, & Pendry, 2000; Prado & Gibbs, 1993). SDTFs and the
Cerrado can be distinguished by physiognomy, function and dissimilarities in phylogenetic
composition (Oliveira-Filho, Pennington, Rotella, & Lavin, 2014; Oliveira-Filho et al., 2013).
The Chaco woodlands at the southwest of the Dry Diagonal are climatically seasonal and its
woodlands do not experience fire. The Chaco woodlands have been considered distinct from
SDTF on the basis that they experience regular frost, greater temperature seasonality and
often distinct edaphic conditions, e.g. hypersaline soils (DRYFLOR, 2017; Prado & Gibbs,
1993). The Pantanal Domain has heterogeneous vegetation including SDTFs, savanna and
swamps, while the Pampa Domain is a largely subtropical grassland that has forest patches
along river courses and on certain edaphic conditions.
Lowland Tropical South America, due to its size, diversity and non-continuous geographic
distribution of biomes and vegetation types, is an ideal system to study how biomes can be
delimited, at a continental scale, through means other than climate and remote sensing. Its
complex environmental controls of both climate and soil point to the necessity of
developing a new approach for biome delimitation that is better linked to biodiversity.
Biome schemes centred on species composition may be more useful for comparative
biology, conservation, and enable a better understanding of the possible mechanistic
relationships between vegetation and environment.
Here we test the utility and performance of a floristic approach for mapping biomes at a
continental scale, with a particular focus on Brazil and neighbouring countries. We use a
dataset of 4,103 geo-referenced floristic inventories of tree species that span the major
climatic and edaphic gradients of the region. We first test how well climatic data perform in
distinguishing among biomes. We hypothesize that climatic data will be able to distinguish
wet forests from the dry biomes, but that it will fail to distinguish SDTF from savanna as they
are often edaphically differentiated (Ratter et al., 1997). We also test the ability of edaphic
data, when considered in conjunction with climate, to increase the accuracy of biome
delimitation. Lastly, we assess how our floristic approach to mapping biomes compares with
the ecoregion-based classification system of Dinerstein et al. (2017) (a revised version of
Olson et al. (2001) system), and then for Brazil only, against the Domain classification of
IBGE (2012). Our use of floristics data may allow for the delimitation and mapping of biomes
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
in a manner directly relevant to managing ecosystems and developing conservation
strategies, for example by enabling the modelling of future climate change effects on
tropical vegetation (Prieto-Torres & Rojas-Soto, 2016; Prieto-Torres et al., 2016).
Methods
The NeoTropTree dataset
Floristic inventories of tree communities were obtained from the NeoTropTree (NTT)
dataset (Oliveira-Filho, 2017), which contains tree species inventories for more than 6,000
geo-referenced sites across South America. Trees are here defined as free-standing woody
plants greater than three metres in height. Every site in the NTT database is based on a tree
species list generated via an inventory, phytosociological survey or floristic survey. These
data sources are derived from published and unpublished literature (e.g. PhD theses,
environmental consultancy reports). Other species are added to the site species list based
on surveys of specimens in herbaria in South America, USA and Europe or online (e.g. CRIA,
2012). All entries are carefully checked for doubtful determinations and synonyms by
consulting the taxonomic literature, the “Flora do Brasil” (http://floradobrasil.jbrj.gov.br/)
and the “Flora del Conosur” (Zuloaga, Belgrano, Zuloaga, & Belgrano, 2015) –
http://www.darwin.edu.ar/), with additional direct consultation of taxonomists. Our data
excludes checklists with < 10 species, because in lowland tropical regions, this is invariably
due to low sampling or collecting efforts, rather than truly low species richness.
The vegetation type for each site, as documented in the original data source, is recorded
and standardized to the vegetation types in Oliveira-Filho (2017; see also Table S1). When a
herbarium voucher of an additional species is noted to come from within a 5 km radius of
the original site, the collection label is checked to ensure that the species is found in the
same vegetation type. Where two or more sites of different vegetation types co-occur
within 10 km (768 sites– 19.13 % of our total), this results in geographically overlapping sites
in the NTT database, each for a distinct vegetation type. Further details of NTT history,
protocols and data can be found at www.neotroptree.info. We restricted analyses to the
tropical and neighbouring subtropical lowlands of South America east of the Andes, and did
not include any NTT site above 1,000 m elevation or below 36o S latitude. Montane areas
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
were excluded because biogeographic barriers may be playing significant roles in floristic
differentiation. Including subtropical sites allowed us to contextualize our results from the
tropics. In total, we included 4,103 individual sites, containing 10,306 tree species from
1,062 genera and 148 families.
Statistical Analyses
We performed hierarchical clustering based on tree species composition to assign sites to
groups in an unsupervised manner (i.e. without reference to any environmental data). For
clustering, we used the Simpson floristic distance amongst sites, which is the complement of
the number of species shared between two sites divided by the maximum number of
species that could be shared between the two sites: 1 - speciesshared/total_speciesminimum
(Baselga, 2010). This is identical to the βsim metric (Kreft & Jetz, 2010), but we use the term
Simpson distance because of its historical precedence (Baselga, 2010). This metric isolates
the effects of species turnover and is not confounded by large differences in species richness
amongst sites (Baselga, 2010). We built 1,000 clusters, each after randomising the row order
in the matrix (species per site), following the procedure of Dapporto et al. (2013). We
removed 24 sites that were unstable in their placement across the 1,000 clusters, which
were identified by co-opting an approach used in phylogenetics to identify ‘rogue taxa’ that
reduce resolution in phylogenetic analyses (Aberer, Krompass, & Stamatakis, 2012). In the
final consensus cluster, only those groups that were present in at least 50% of the clusters
are distinguished (Omland, Cook, & Crisp, 2008). This analysis was performed in R (R Team,
2016) using the “recluster” package (Dapporto et al., 2015).
To determine the biome identity of clusters, we used a reciprocal illumination procedure of
assessing the overall structure of the cluster while considering site vegetation types (see
Table S1). This process is inherently fractal and one could identify increasingly smaller
groups of sites. We focused on defining biomes in the broadest sense in order to increase
their generality and utility, and our delimitations were performed in the context of the main
biomes that have previously been proposed for LTSA, namely wet or moist tropical forests
(hereafter wet forests), SDTF, subtropical forests, savanna and chaco woodlands. In essence,
our approach tested if there is floristic integrity to these previously proposed biomes, and
we found clear evidence that there was, i.e. higher-level groups were comprised largely of
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
one broad biome type (Table S1). For heuristic purposes, we constructed a continuous
biome map by applying Thiessen’s polygons method in ArcGIS 10.4.1 (ESRI, 2017). This
approach expands a polygon of a given biome classification for each NTT site until the
polygons from neighbouring NTT sites are encountered. If they represent the same biome,
then the polygons are fused and this procedure is continued until the entirety of the study
area was categorised to biome.
We assessed which sites may be intermediate or transitional between our biomes using a
silhouette analysis, via the R package cluster (Maechler, Rousseeuw, Struyf, Hubert, &
Hornik, 2016). We also visually assessed where these ambiguously classified sites are
located in species compositional space by means of a non-metric multidimensional scaling
analysis (NMDS, McCune & Grace, 2002) of sites in two dimensions based on the Simpson
distance amongst sites.
Using climate and edaphic data to distinguish biomes
To assess if the biomes identified could be distinguished using climatic data, with or without
edaphic data, we used a Random Forest classification tree approach (Breiman, 2001),
implemented in the randomForest package in the R Statistical Software (Liaw & Wiener,
2002). We used 19 bioclimatic variables developed by (Hijmans, Cameron, Parra, Jones, &
Jarvis, 2005), which quantify various aspects of temperature and precipitation regimes, as
well as an estimate of average maximum climatological water deficit (CWD) per year (Chave
et al., 2014). As edaphic variables, we included pH (extracted with KCl), cation exchange
capacity (cmol/kg) and percentage of sand, silt and clay extracted from SoilGrids v0.5.5
(https://soilgrids.org/, (Hengl et al., 2017) at four different soil depths: 0 cm, 5 cm, 15 cm
and 30 cm, which were then averaged. Two different classifications were performed, one
considering climatic data alone and another considering both climatic and soil data.
In order to assess the success rate of the classification tree approach in assigning sites to
biome and to determine which biomes were incorrectly classified, we generated confusion
matrices, which show assignment based on climate alone or climate and soil versus
assignment done above based on vegetation type and tree species composition. We also
estimated the importance of each variable for distinguishing biomes using Breiman’s
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
measure of importance (Breiman, 2001). As we had substantial variation in sample size
amongst our biomes that could bias importance measures, we equalized the number of sites
across all biomes by rarefying to the number of sites present in the most poorly sampled
biome. Rarefactions were performed randomly 100 times and variable importance values
were averaged across the 100 replicates. In order to understand climatic overlaps amongst
biomes, we additionally plotted sites in a pairwise manner for key climatic variables (MAP,
MAT and CWD).
Comparison to existing biome maps
We compared how two commonly used vegetation maps for South America classify sites to
biome compared to our analyses. We focused on the map of Dinerstein et al. (2017), in
which ecoregions are grouped into biomes and which is a revised version of Olson et al.
(2001), and the Brazilian Domain system (IBGE 2012). We determined which biomes and
domains in these systems conceptually correspond to the biomes we established here, and
assessed how often these mapping systems gave the same identity to our NTT sites. The
ecoregion data layer was obtained from https://ecoregions2017.appspot.com/ and the IBGE
Domain data layer from http://www.geoservicos.ibge.gov.br/geoserver/web/ (layer
CREN:biomas_5000).
Results
Biomes of Lowland Tropical South America
Hierarchical cluster analysis produced five higher-level groups (Fig. 1), which we designated
as biomes based on a priori vegetation type classifications. Wet forests fell into two
different groups, which we tentatively treat as separate biomes. One comprises sites in the
Amazon and the Guiana Shield, which we refer to as the Amazon Forest biome, and the
other is comprised of sites along the Atlantic coast, which we refer to as the Atlantic Forest
biome (Fig. 2). These two biomes are largely concordant with the Amazon and Atlantic
Forest Domains, except that they also include semideciduous and gallery forests, found well
outside of the geographic areas of the forest Domains (Fig. 2).
The other three major groups in the cluster are found primarily in the Dry Diagonal, which
extends from northeast Brazil to Bolivia, Paraguay and northern Argentina (Fig. 2). One,
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
which we refer to as Savanna, comprises sites with a grassy understorey found throughout
central Brazil and eastern Bolivia, overlapping with the Cerrado Domain, but with disjunct
occurrences in the Amazon Forest and Atlantic Forest biomes. The Savanna biome is clearly
distinguished floristically from a biome that we term Seasonally Dry Tropical Forest (SDTF),
based on the original vegetation classifications of sites (Table S1). The SDTF biome has a
discontinuous distribution from the Pantanal and Chiquitania in Bolivia and southern Brazil
to its largest extension in the Caatinga Domain of northeastern Brazil (Fig. 2). It is spatially
interdigitated with the Savanna biome. The last group, which we distinguish as a separate
biome is the Chaco, comprising woodlands in Bolivia, Argentina and Paraguay and extending
to the borders of southern Brazil. While most of the sites in the Chaco biome cluster are
subtropical and experience frost, there are a significant number of sites found north of 23
degrees latitude that are unlikely to experience freezing and can be considered tropical (Fig.
2). See Supplementary Materials (Appendix 1) for further description of the biomes. Our
continuous biome map, developed using the Thiessen’s polygons method, shows the LTSA
biomes’ overall spatial distribution and highlights the regions in which they interdigitate
(Figure 3).
Of 4,103 sites, 1,097 were classified as Amazon Forest, 1,566 as Atlantic Forest, 760 as
Savanna, 564 as SDTF and 116 as Chaco. Silhouette analysis (Fig S1) showed that 271 sites
are floristically more similar to a different biome than that with which they were original
clustered, which we interpret to indicate that these sites are transitional between two
biomes (Fig. 4a, Table S2). An ordination of sites (NMDS with two axes, stress value= 0.1816)
also suggests that these sites are compositionally transitional (Fig. 4b). Floristically
transitional sites were common between the Amazon and Atlantic Forest biomes (53 sites),
between the Savanna and Atlantic Forest biomes (115 sites), and between the SDTF and
Atlantic Forest biomes (49 sites), while they were infrequent between other biomes,
including between any pair of dry biomes. Floristically transitional sites are common in the
Dry Diagonal (Fig. 4a), particularly between the Cerrado and the Amazon Forest and
between the Chaco and the Atlantic Forest. Many of the gallery forests within the Cerrado
Domain also have an ambiguous tree species compositional identity and are therefore
difficult to classify.
Using climate and edaphic data to distinguish biomes
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
We find that biomes overlap substantially in climatic space, both in terms of water
availability (Fig. 5) and temperature (Fig. 6). For example, all five biomes defined here
occupy at least two of the climatic biomes proposed by Whittaker (1975) (Fig. 6). Of the
3,832 sites that are not considered transitional in nature, 712 were misclassified based on
climate (18.6% of sites; Table 1). Considering all sites together, including transitional ones,
we found a slightly higher error rate of 20.7% (Table S3). The most common
misclassifications involved Amazon or Atlantic Forest sites being classified as belonging to
the Savanna biome or vice versa, while climatic misclassifications of SDTF and Savanna were
also common (Table 1). Sites in the Amazon and Atlantic Forest wet biomes were distinct
climatically. Meanwhile, the Chaco biome was rarely confused climatically with any of the
other biomes. These patterns did not change when sites that have centres within 10 km of
each other, i.e. overlapping in geographic space, were removed (Table S4, error rate:
20.3%).
The inclusion of edaphic variables slightly increased overall classification success by 3.2%
(Table 2), and 3% when transitional sites were included (Table S5). There were a total of 124
sites that switched from being classified incorrectly (with just climatic data) to being
classified correctly (once edaphic data were included; Table 2). Most of these were Savanna
sites classified as Atlantic Forest and vice-versa.
Whether or not edaphic variables are included, the three main most important variables for
classification were Mean Annual Precipitation, Temperature Seasonality and Maximum
Climatological Water Deficit (Table 3). Overall, climatic variables seem to be more important
than edaphic variables for distinguishing biomes, with variables related to precipitation,
water availability and temperature seasonality ranking higher than variables related to
mean temperature. However, overall we do have fewer edaphic variables and pH and cation
exchange capacity (CEC) are among the top 10 variables (Table 3).
Comparison to existing biome maps
The classification systems developed by Olson and Dinerstein et al. (2001, 2017) and IBGE
(2012) assigned 74-75% of the NTT sites to the same biomes as they were placed according
to our analyses (74.7% Dinerstein et al., 2017, Table S6; 74.5% IBGE, 2012, Table S7). In
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
Dinerstein’s system, the majority of the misclassification results from Atlantic Forest sites
being incorrectly classified as Tropical or Subtropical Savannas and Savanna being classified
as Tropical Moist Forest (Figure S2). In IBGE’s system, the error rate stems from SDTF sites
being classified as Cerrado and vice-versa (Figure S3).
Discussion
Our study demonstrates that using climatic data alone, with or without supplementary
edaphic data, to map biomes would result in substantial error, causing misclassification of
15.2 - 20.7% of sites. Such misclassifications are due to pronounced climatic overlap of
biomes (Figs 5, 6) and to edaphic heterogeneity at small spatial scales that is not captured
by available data, which are derived via interpolation among relatively sparse soil sampling.
Recently, researchers have begun assigning study sites to biomes, generally those of
Whittaker (1975) based solely on climatic values, e.g. mean annual precipitation and
temperature (e.g. Díaz et al., 2016; Qian & Ricklefs, 2017; Siepielsky et al., 2017). Our results
suggest this is potentially problematic (Fig. 6). For example, the Amazon and Atlantic Forests
can both occur in areas that are more seasonal than ‘tropical rain forest’ (sensu Whittaker),
while the Savanna biome can occur in much wetter areas than indicated by Whittaker (1975;
see also Lehmann et al., 2014). It is notable that none of our five major biomes are
restricted to a single biome in Whittaker’s climatic biome classification (Fig. 6).
We were able to employ a floristic approach to mapping biomes at a continental scale.
Recent biome maps of LTSA, generally based on remote sensing, either fail to include major
biomes (e.g. Seasonally Dry Tropical Forest is absent from Hirota et al., 2011; Staver et al.,
2011), or are unable to distinguish amongst the dry tropical biomes of Savanna and SDTF
(Beuchle et al., 2015). While floristic approaches to mapping biomes are unlikely to succeed
inter-continentally because of the lack of shared species or even genera at this scale (Dexter
et al., 2015), the increasing availability of floristic composition and species distribution
information (e.g. www.gbif.org, www.forestplots.net, www.neotroptree.info) should allow
this approach to be implemented within continents. It is important to note that any
complete and continuous (or ‘wall-to-wall’) map of biome distribution will be inaccurate at
small spatial scales due to high edaphic and floristic heterogeneity coupled with incomplete
sampling. We have generated a continuous map (Fig. 3), but its purpose is as a heuristic
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
scheme to understand patterns in the distribution of biomes in LTSA. We do not contend
that every point on the map is accurately classified, as that would belie one of the principal
outcomes of this study, that of high biome heterogeneity at small spatial scales, as
previously noted by Pennington et al. (2006), Werneck (2011), Collevatti et al. (2013).
Biomes of Lowland Tropical South America
Our analyses suggest three to five major biomes in LTSA. The Amazon and Atlantic Forests
might represent separate biomes, whereas previously they have often been considered as a
single tropical wet/moist forest biome. They are floristically distinct and their climatic niches
are almost completely non-overlapping. Our floristic circumscription of the Atlantic Forest
matches the sensu-latissimo definition of Oliveira-Filho, Jarenkow, & Rodal (2006). Our
delimitation of the Amazon Forest is similar to previous studies that include the majority of
the Amazon Basin drainage and the Guianan Shield (e.g., Prance, 1982; ter Steege et al.,
2006), although we note that our sampling of the Guianan Shield is limited.
The Savanna biome is floristically distinct from the other dry biomes, which is expected
since it is a uniquely disturbance driven system, strongly influenced by fire (Archibald,
Lehmann, Gómez-dans, & Bradstock, 2013; Ratter et al., 1997). Many sites in the SDTF
biome are often drier, in terms of MAP and CWD, than the majority of sites in the Savanna
biome (Fig. 5), which runs counter to thinking that tropical wet forest transitions to tropical
seasonal forest and then to savanna as water availability declines (e.g. Malhi et al., 2009).
Meanwhile, our results from floristic analyses give support to previous studies (DRYFLOR,
2016; Pennington, Lavin, & Oliveira-Filho, 2009; Pennington et al., 2000; Prado & Gibbs,
1993) that have argued that the SDTFs scattered across lowland tropical South America
should be regarded as a single biome, with the exclusion of the Chaco. We find that the
climatic niches of Chaco and SDTF do not overlap, with the Chaco occurring in a colder
climate with much higher temperature seasonality. However, further studies are needed
that compare ecosystem function in the Amazon versus Atlantic Forests and in the SDTF
versus Chaco to verify their status as distinct biomes. For further discussion of floristic
patterns within and across biomes, please refer to the supplementary material (Appendix 1).
Using climate and edaphic data to distinguish biomes
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
Mean annual precipitation (MAP), several measures of dry season precipitation and water
deficit, temperature variability and soil pH were the most important environmental
variables in distinguishing major biomes (Table 2). That precipitation-related variables are
on average more important than temperature-related variables is to be expected, given that
the majority of our sampling and most of the biomes under study are within the tropics, and
thus represent a limited range of non-freezing temperature regimes (Augusto, Davies,
Delzon, De Schrijver, & Chave, 2014). Nevertheless, it is notable that measures of
temperature variability, particularly across seasons, were more important than other
temperature measures, including mean annual temperature (MAT) and minimum
temperature of the coldest month. This may be because plant species’ ranges are often
constrained by how much temperature can vary in a given location, and by temperature
extremes (O’Sullivan et al., 2017).
While a classification success rate of 80% seems high, this would result in 1 in 5 sites being
misclassified, which is potentially problematic for conservation and management decisions.
Some sites are floristically transitional in nature and inherently difficult to classify. Such
transitional sites may be particularly resilient, and thus important, under future climate
change, and they may require their own management regimes (Prieto-Torres et al. 2016).
Regardless, the high error rate (18.6%) among non-transitional sites (sites not detected by
the silhouette analysis as belonging to a different biome) is still of concern as they comprise
93.4% of our sites. In order to improve classification of these sites to biome based on
environmental data, environmental data in better resolution are needed. Publicly available
environmental data are derived from interpolation. For climate, which varies at a relatively
broad spatial grain, this may not be problematic. However, edaphic data vary at a small
spatial grain, and interpolation-based methods may be inadequate to capture edaphic
conditions at many sites. Also, the edaphic data from SoilGrids does not include variables,
such as soil fertility (sum of bases), phosphorous and aluminium content, which are highly
relevant to tree species growth. Meanwhile, other non-climatic and non-edaphic variables,
such as fire and disturbance, may play a significant role in determining tree species
composition at local sites, and biome identity more widely. For example, SDTF and wet
forest can convert to savanna if there is sufficient disturbance via fire or anthropogenic
woody biomass removal (Devisscher, Anderson, Aragão, Galván, & Malhi, 2016).
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
Comparing to existing biome maps
The comparisons between the classification system presented here and those of Dinerstein
et al. (2017) and the Domain system (IBGE, 2012) revealed a ~25% misclassification rate for
the latter two. These high error rates stem from two sources: the intrusion of SDTF and the
Atlantic and Amazon Forests (as gallery forest) into the Savanna biome in the dry diagonal,
and the existence of non-equivalent categories among these systems. Dinerstein et al.
(2017) and IBGE (2012) recognize tropical and subtropical wetlands (named Pantanal in
IBGE’s system) as a distinct biome or domain, while the IBGE Domain system also delimits
the Pampas (a.k.a. Campos Sulinos - southern Brazilian steppes). These two categories have
not been detected and classified by our approach. Rather, the region classified as Pantanal
by IBGE (2012) is covered by a mix of different vegetation formations that are floristically
similar to SDTFs, Savannas and also the semideciduous portion of the Atlantic Forest. The
forests within the area known as the Pampas at South Brazil are floristically similar, in
relation to tree species composition, to the rest of the subtropical portion of the Atlantic
Forest biome (Oliveira-Filho, Budke, Jarenkow, Eisenlohr, & Neves, 2015).
Synergies between biodiversity conservation and ecosystem management
Delimiting biomes based on tree species composition offers the possibility of synergy
between ecosystem management planning and conservation prioritisation. The biomes we
have delimited differ in tree species composition and therefore likely differ in ecosystem
function. Ecosystem management plans should therefore be developed separately for each.
Similarly, these biomes have almost no species in common, yet have many species unique to
them. Our schematic map (Fig. 3) also indicates how these biomes are distributed at a
continental scale, highlighting how discontinuous biome distribution can be in LTSA. These
are important observations that must be considered in conservation and management. As
an example, it is only recently that the SDTF have been recognised as a biome (Gentry,
1995; Murphy & Lugo, 1986; Prado & Gibbs, 1993), a definition consistent with our analyses,
and there is no synthetic conservation plan that addresses the biome as a whole across the
Neotropics (though see DRYFLOR 2016 for first steps). Current conservation planning for
SDTF in Brazil focuses solely on the Caatinga Domain, but many Brazilian SDTFs are found in
disjunct patches outside of this area, especially in the Cerrado, placing them under laws
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
designed to protect savanna diversity. As another example, the Chaco is under great threat
due to an increase of habitat destruction and fragmentation during the last 30 years
(Hansen et al., 2013, Nori et al. 2016), but if recognised as a separate biome, as our analyses
suggest, the urgency of its conservation may be better recognised (Kuemmerle et al., 2017).
Conclusions
We have mapped the principal biomes in LTSA by using information on tree species
composition of > 4,000 sites. The Savanna, Amazon and Atlantic Forest and SDTF biomes
have an interdigitated distribution in central South America and overlap substantially in
climatic space. Biome distribution cannot therefore be fully accounted for by climate,
suggesting that climate projections alone will be insufficient to predict future biome shifts.
Additional, meaningful environmental variables (e.g. available nitrogen, phosphorous,
aluminium, etc.) must be measured and accounted for in models. The interdigitiation of
biomes, especially in the dry diagonal across Brazil, is not recognised in the current IBGE
(2012) system on which Brazilian conservation legislation is based, leading to the neglect of
highly threaten SDTF vegetation outside of the Caatinga Domain. Our analyses also show
Chaco and SDTF are distinct, which must be considered in land management and
conservation. We suggest that species composition can be central to delimiting meaningful
biomes for comparative research and conservation.
Acknowledgements
P.L.S.M. thanks the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brazil
(CAPES) for supporting a full PhD at the University of Edinburgh under the Science without
Borders Programme (grant 99999.013197/2013-04). A.O.F. was supported by the Conselho
Nacional de Desenvolvimento Científico e Tecnológico—Brazil (CNPq) (grant 301644/88-8).
D.M.N., R.T.P. and K.G.D. were supported by the National Environment Research Council
(grant NE/I028122/1). P.L.S.M also thanks Chrystiann Lavarini for help in making Figure 3.
Data Accessibility
The data used to produce this paper can be freely accessed at
http://www.neotroptree.info/.
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
Biosketch
Pedro L. Silva de Miranda is a PhD student at the University of Edinburgh under the
supervision of Dr. Kyle G. Dexter, Dr. Toby Pennington, Dr. Caroline Lehmann and Professor
Ary Oliveira-Filho. His PhD focuses on identifying the main biomes in Lowland Tropical South
America, understanding the main environmental drivers behind them (climatic and edaphic)
and assessing the effects climate change will have on their distribution and diversity. The
research group – PLant Evolutionary Ecologists and BiogeographerS, PLEEBS
(http://phylodiversity.net/kdexter/HOME.html) – is led by Dr. Kyle Dexter and focuses
primarily on tropical vegetation.
References
Aberer, A. J., Krompass, D., & Stamatakis, A. (2012). Pruning Rogue Taxa Improves Phylogenetic Accuracy. Systematic Biology. 62, 162–166.
Archibald, S., Lehmann, C. E. R., Gómez-dans, J. L., & Bradstock, R. A. (2013). Defining pyromes and global syndromes of fire regimes. Proceedings of the National Academy of Sciences, 110, 6442–6447.
Augusto, L., Davies, T. J., Delzon, S., De Schrijver, A., & Chave, J. (2014). The enigma of the rise of angiosperms: can we untie the knot?. Ecology Letters, 17, 1326-1338.
Baker, T. R., Pennington, R. T., Dexter, K. G., Fine, P. V. A., Fortune-Hopkins, H., Honorio, E. N., … Vasquez, R. (2017). Maximising Synergy among Tropical Plant Systematists, Ecologists, and Evolutionary Biologists. Trends in Ecology & Evolution, 32, 258–267.
Baselga, A. (2010). Partitioning the turnover and nestedness components of beta diversity. Global Ecology and Biogeography, 19, 134–143.
Beuchle, R., Grecchi, R. C., Shimabukuro, Y. E., Seliger, R., Eva, H. D., Sano, E., & Achard, F. (2015). Land cover changes in the Brazilian Cerrado and Caatinga biomes from 1990 to 2010 based on a systematic remote sensing sampling approach. Applied Geography, 58, 116–127.
Box, E. O. (1995). Factors determining distributions of tree species and plant functional types. Vegetatio, 121, 101–116.
Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32.
Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M. S., Delitti, W. B. C., … Vieilledent, G. (2014). Improved allometric models to estimate the aboveground biomass of tropical trees. Global Change Biology, 20, 3177–3190.
Clements, F. E. (1916). The Development and Structure of Biotic Communities. In Ecological Society of America (Vol. 5, pp. 119–128). New York.
499
500
501
502
503
504
505
506
507
508
509510
511512
513514
515516517
518519
520521522
523524
525
526527528
529530
Collevatti, R. G., Lima-Ribeiro, M. S., Diniz-Filho, J. A. F., Oliveira, G., Dobrovolski, R., & Terribile, L. C. (2013). Stability of Brazilian Seasonally Dry Forests under Climate Change: Inferences for Long-Term Conservation. American Journal of Plant Sciences, 04, 792–805.
CRIA. (2012). speciesLink. Retrieved from: http://www.splink.org.br/index
Dapporto, L., Ramazzotti, M., Fattorini, S., Talavera, G., Vila, R., & Dennis, R. L. H. (2015). Package ‘recluster’ - Ordination methods for the analysis of beta-diversity indices. Retrieved from: https://cran.r-project.org/web/packages/recluster/recluster.pdf
Dapporto, L., Ramazzotti, M., Fattorini, S., Talavera, G., Vila, R., & Dennis, R. L. H. (2013). Recluster: An unbiased clustering procedure for beta-diversity turnover. Ecography, 36, 1070–1075.
Devisscher, T., Anderson, L. O., Aragão, L. E. O. C., Galván, L., & Malhi, Y. (2016). Increased Wildfire Risk Driven by Climate and Development Interactions in the Bolivian Chiquitania, Southern Amazonia. PLoS ONE, 11(9): e0161323.
Dexter, K. G., Smart, B., Baldauf, C., Baker, T. R., ..., Pennington, R. T. (2015). Vegetation in seasonally dry regions of the tropics: floristics and biogeography. International Forestry Review, 17, 10–32.
Díaz, S., Kattge, J., Cornelissen, J. H. C., Wright, I. J., Lavorel, S., Dray, S., … Gorné, L. D. (2016). The global spectrum of plant form and function. Nature, 529, 167–171.
Dinerstein, E., Olson, D., Joshi, A., Vynne, C., Burgess, N. D., Wikramanayake, E., … Saleem, M. (2017). An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm. BioScience, 67, 534–545.
DRYFLOR. (2016). Plant diversity patterns in neotropical dry forests and their conservation implications. Science, 353, 1383–1387.
DRYFLOR, Pennington, R. T., Banda-R, K., Delgado-Salinas, A., Dexter, K. G., Galetti, L., … Särkinen, T. (2017). Response - Forest conservation: Remember Gran Chaco. Science, 355, 465–466.
Engemann, K., Sandel, B., Enquist, B. J., JØrgensen, P. Mø., Kraft, N., Marcuse-Kubitza, A., … Svenning, J. C. (2016). Patterns and drivers of plant functional group dominance across the Western Hemisphere: A macroecological re-assessment based on a massive botanical dataset. Botanical Journal of the Linnean Society, 180, 141–160.
ESRI. (2017). ArcGIS 10.4.1. Redlands: CA: Environmental Systems Research Institute.
Friedl, M. A., McIver, D. K., Hodges, J. C., Zhang, X. Y., Muchoney, D., & Strahler. (2002). Global land cover mapping from MODIS: algorithms and early results. Remote Sensing of Environment, 83, 287–302.
Gentry, A. H. (1995). Patterns of diversity and floristic composition in Neotropical montane forest. In S. P. Churchill, H. Balslev, E. Forero, & J. L. Luteyn (Eds.), Biodiversity and Conservation of Neotropical Montane Forest (pp. 103–126). New York: The New York Botanical Garden.
Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A., … Townshend, J. R. G. (2013). High-Resolution Global Maps of 21st-Century Forest Cover Change. Science, 342, 850–853.
531532533
534
535536537
538539
540541542
543544
545546
547548549
550551
552553
554555556557
558
559560561
562563564
565566567
Hengl, T., Jesus, J. M. de, Heuvelink, G. B. M., Gonzalez, M. R., Kilibarda, M., Blagotić, A., … Kempen, B. (2017). SoilGrids250m: Global gridded soil information based on machine learning. PLoS ONE, 12(2): e0169748.
Higgins, S. I., Buitenwerf, R., & Moncrieff, G. R. (2016). Defining functional biomes and monitoring their change globally. Global Change Biology, 22, 3583–3593.
Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 1965–1978.
Hirota, M., Holmgren, M., Van Nes, E. H., & Scheffer, M. (2011). Global Resilience of Tropical Forest and Savanna to Critical Transitions. Science, 334, 232–235.
Holdridge, L. R. (1947). Determination of world plant formations from simple climatic data. Science, 105, 267–268.
Humboldt, C. A. (1816). XCIII. On the laws observed in the distribution of vegetable forms. The Philosophical Magazine, 47, 446–453.
Instituto Brasileiro de Geografia e Estatística (Ed.). (2012). Manual técnico da vegetação brasileira (2a̲ edição revista e ampliada). Rio de Janeiro: Instituto Brasileiro de Geografia e Estatística-IBGE.
Kreft, H., & Jetz, W. (2010). A framework for delineating biogeographical regions based on species distributions. Journal of Biogeography, 37, 2029–2053.
Kuemmerle, T., Baldi, G., Camino, M., Cuellar, E., Cuellar, R. L., Decarre, J., … Semper-pascual, A. (2017). Forest conservation : Remember Gran Chaco Forest conservation: Humans’ handprints. Science, 355, 465.
Lamanna, C., Blonder, B., Violle, C., Kraft, N. J. B., Sandel, B., Šímová, I., … Enquist, B. J. (2014). Functional trait space and the latitudinal diversity gradient. Procedings of the National Academy of Sciences, 111, 13745–13750.
Lehmann, C. E. R., Anderson, T. M., Sankaran, M., Higgins, S. I., Archibald, S., Hoffmann, W. A., … Bond, W. J. (2014). Savanna Vegetation-Fire-Climate Relationships Differ Among Continents. Science, 343, 548–553.
Liaw, A., & Wiener, M. (2002). Classification and Regression by randomForest. R News, 2, 18–22.
Maechler, M., Rousseeuw, P., Struyf, A., Hubert, M., & Hornik, K. (2016). cluster: Cluster Analysis Basics and Extensions.
Malhi, Y., Aragao, L. E. O. C., Galbraith, D., Huntingford, C., Fisher, R., Zelazowski, P., … Meir, P. (2009). Exploring the likelihood and mechanism of a climate-change-induced dieback of the Amazon rainforest. Proceedings of the National Academy of Sciences, 106, 20610–20615.
McCune, Bruce, James B. Grace, and Dean L. Urban. Analysis of ecological communities. Vol. 28. Gleneden Beach, OR: MjM software design, 2002.
Mittermeier, R. A., Gil, P. G., Hoffmann, M., Pilgrim, J., Brooks, T., Mittermeier, C. G., … Fonseca, G. A. B. (2004). Hotspots Revisited (1st ed.). Mexico City: CEMEX.
568569570
571572
573574575
576577
578579
580581
582583
584585
586587588
589590591
592593594
595
596597
598599600
601602
603604
Mittermeier, R. A., Myers, N., Thomsen, J. B., da Fonseca, G. A. B., & Olivieri, S. (1998). Biodiversity Hotspots and Major Tropical Wilderness Areas: Approaches to Setting Conservation Priorities. Conservation Biology, 12, 516–520.
Moncrieff, G. R., Hickler, T., & Higgins, S. I. (2015). Intercontinental divergence in the climate envelope of major plant biomes. Global Ecology and Biogeography, 24, 324–334.
Murphy, P. G., & Lugo, A. E. (1986). Ecology of Tropical Dry Forest. Annual Review of Ecology and Systematics, 17, 67–88.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., Fonseca, G. A. B., & Kent, J. (2000). Biodiversity hotspots for conservation priorities. Nature, 403, 853–858.
Neves, D. M., Dexter, K. G., Pennington, R. T., Bueno, M. L., & Oliveira Filho, A. T. (2015). Environmental and historical controls of floristic composition across the South American Dry Diagonal. Journal of Biogeography, 42, 1566–1576.
Nori, J., Torres, R., Lescano, J. N., Cordier, J. M., Periago, M. E., Baldo, D. & Di Minin, E. (2016). Protected areas and spatial conservation priorities for endemic vertebrates of the Gran Chaco, one of the most threatened ecoregions of the world. Diversity and Distributions, 22, 1212-1219.
Odum, E. P. (1975). Ecology (2nd ed.). London: Holt Rinehart and Winston.
Oliveira-Filho, A. T. (2017). NeoTropTree, Flora arbórea da Região Neotropical: Um banco de dados envolvendo biogeografia, diversidade e conservação. Retrieved from http://www.neotroptree.info/
Oliveira-Filho, A. T., Budke, J. C., Jarenkow, J. A., Eisenlohr, P. V., & Neves, D. R. M. (2015). Delving into the variations in tree species composition and richness across South American subtropical Atlantic and Pampean forests. Journal of Plant Ecology, 8, 242–260.
Oliveira-Filho, A. T., Cardoso, D., Schrire, B. D., Lewis, G. P., Pennington, R. T., Brummer, T. J., … Lavin, M. (2013). Stability structures tropical woody plant diversity more than seasonality: Insights into the ecology of high legume-succulent-plant biodiversity. South African Journal of Botany, 89, 42–57.
Oliveira-Filho, A. T., Jarenkow, L. A., & Rodal, M. J. N. (2006). Floristic relationships of seasonally dry forests of Eastern South America based on tree species distribution patterns. In Neotropical Savannas and Seasonally Dry Forests: Plant Diversity, Biogeography, and Conservation (pp. 159–192). London: Taylor & Francis.
Oliveira-Filho, A. T., Pennington, R. T., Rotella, J., & Lavin, M. (2014). Exploring evolutionarily meaningful vegetation definitions in the tropics: a community phylogenetic approach. In D. A. Coomes, D. F. R. P. Burslem, & W. D. Simonson (Eds.), Forests and Global Change (pp. 239–260). Cambridge: Cambridge University Press.
Oliveira-Filho, A. T., & Ratter, J. A. (1995). A study of the origin of central brazilian forests by the analysis of plant species distribution patterns. Edinburgh Jounal of Botany, 52, 141–194.
Olson, D. M., & Dinerstein, E. (1998). The Global 200: A Representation Approach to Conserving the Earth’s Most Biologically Valuable Ecoregions. Conservation Biology, 12, 502–515.
Olson, D. M., Dinerstein, E., Wikramanayake, E. D., Burgess, N. D., Powell, G. V. N., Underwood, E. C., … Kassem, K. R. (2001). Terrestrial Ecoregions of the World: A New Map of Life on EarthA new global
605606607
608609
610611
612613
614615616
617618619
620
621622
623624625
626627628629
630631632633
634635636637
638639
640641
642643
map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience, 51, 933–938.
Omland, K. E., Cook, L. G., & Crisp, M. D. (2008). Tree thinking for all biology: the problem with reading phylogenies as ladders of progress. BioEssays : News and Reviews in Molecular, Cellular and Developmental Biology, 30, 854–867.
O’Sullivan, O. S., Heskel, M. A., Reich, P. B., Tjoelker, M. G., Weerasinghe, L. K., Penillard, A., … Atkin, O. K. (2017). Thermal limits of leaf metabolism across biomes. Global Change Biology, 23, 209–223.
Pennington, R. T., Lavin, M., & Oliveira-Filho, A. (2009). Woody Plant Diversity, Evolution, and Ecology in the Tropics: Perspectives from Seasonally Dry Tropical Forests. Annual Review of Ecology, Evolution, and Systematics, 40, 437–457.
Pennington, R. T., Prado, D. E., & Pendry, C. A. (2000). Neotropical seasonally dry forests and Quaternary vegetation changes. Journal of Biogeography, 27, 261–273.
Prado, D. E., & Gibbs, P. E. (1993). Patterns of Species Distributions in the Dry Seasonal Forests of South America. Annals of the Missouri Botanical Garden, 80, 902–927.
Prance, G. T. (1982). Forest Refuges: evidence from woody angiosperms. In G. T. Prance (Ed.), Biological diversification in the tropics (pp. 137-156). New York, NY: Columbia University Press.
Prentice, I. C., Cramer, W., Harrison, S. P., Leemans, R., Robert, A., Solomon, A. M., … Solomon, A. M. (1992). A Global Biome Model Based on Plant Physiology and Dominance , Soil Properties and Climate. Journal of Biogeography, 19, 117–134.
Prentice, K. C. (1990). Bioclimatic distribution of vegetation for general circulation models. Journal of Geophysical Research, 95, 11811–11830.
Prieto-Torres, D. A., & Rojas-Soto, O. R. (2016). Reconstructing the Mexican Tropical Dry Forests via an Autoecological Niche Approach: Reconsidering the Ecosystem Boundaries. PLOS ONE, 11.
Prieto-Torres, D. A., Navarro-Siguenza, A. G., Santiago-Alarcon, D. & Rojas-Soto, O. R. (2016) Response of the endangered tropical dry forests to climate change and the role of Mexican Protected Areas for their conservation. Global Change Biology, 22, 364-379.
Qian, H., Jin, Y., & Ricklefs, R. E. (2017). Patterns of phylogenetic relatedness of angiosperm woody plants across biomes and life-history stages. Journal of Biogeography, 44, 1383-1392.
Ratter, J. A., Ribeiro, J. F., & Bridgewater, S. (1997). The Brazilian Cerrado Vegetation and Threats to its Biodiversity. Annals of Botany, 80, 223–230.
Sandel, B., Gutiérrez, A. G., Reich, P. B., Schrodt, F., Dickie, J., & Kattge, J. (2015). Estimating the missing species bias in plant trait measurements. Journal of Vegetation Science, 26, 828–838.
Särkinen, T., Iganci, J. R. V, Linares-Palomino, R., Simon, M. F., & Prado, D. E. (2011). Forgotten forests - issues and prospects in biome mapping using Seasonally Dry Tropical Forests as a case study. BMC Ecology, 11, 1-15.
Schimper, A. F. W. (1903). Plant-geography upon a physiological basis. Oxford: Clarendon Press.
644645
646647648
649650
651652653
654655
656657
658659
660661662
663664
665666
667668669
670671
672673
674675
676677678
679
Siepielski, A. M., Morrissey, M. B., Buoro, M., Carlson, S. M., Caruso, C. M., Clegg, S. M., … Maccoll, A. D. C. (2017). Precipitation drives global variation in natural selection Adam. Science, 355, 959–962.
Staver, a C., Archibald, S., & Levin, S. A. (2011). The global extent and determinants of savanna and forest as alternative biome states. Science, 334, 230–232.
R Core Team (2017). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
ter Steege, H., Pitman, N. C. A., Phillips, O. L., Chave, J., Sabatier, D., Duque, A., … Vásquez, R. (2006). Continental-scale patterns of canopy tree composition and function across Amazonia. Nature, 443, 444–447.
Van Bodegom, P. M., Douma, J. C., & Verheijen, L. M. (2014). A fully traits-based approach to modeling global vegetation distribution. Proceedings of the National Academy of Sciences of the United States of America, 111, 13733–13738.
Vanzolini, P. E. (1963). Problemas faunísticos do cerrado. Simpósio Sobre o Cerrado, 307, 1.
Veloso, H. P., Rangel Filho, A. L. R., & Lima, J. C. A. (1991). Classificação da vegetação brasileira, adaptada a um sistema universal. Rio de Janeiro: Ministério da Economia, Fazenda e Planejamento, Fundação Instituto Brasileiro de Geografia e Estatística, Diretoria de Geociências, Departamento de Recursos Naturais e Estudos Ambientais.
Violle, C., Borgy, B., & Choler, P. (2015). Trait databases: Misuses and precautions. Journal of Vegetation Science, 26, 826–827.
Violle, C., Reich, P. B., Pacala, S. W., Enquist, B. J., & Kattge, J. (2014). The emergence and promise of functional biogeography. Proceedings of the National Academy of Sciences, 111, 13690–13696.
Walter, H. (1973). Vegetation of the earth in relation to climate and the eco-physiological conditions. New York: Springer.
Whittaker, R. H. (1975). Community and ecosystems. New York: McMillan.
Whittaker, R. J., Araújo, M. B., Jepson, P., Ladle, R. J., James, E., Watson, M., … Willis, K. J. (2005). Conservation Biogeography : Assessment and Prospect. Diversity and Distributions, 11, 3–23.
Woodward, F. I., Lomas, M. R., & Kelly, C. K. (2004). Global climate and the distribution of plant biomes. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 359, 1465–76.
Zuloaga, F. O., Belgrano, M. J. (2015). The Catalogue of Vascular Plants of the Southern Cone and the Flora of Argentina: their contribution to the World Flora. Rodriguésia, 66, 989–1024.
680681
682683
684685
686687688
689690691
692
693694695696
697698
699700
701702
703
704705
706707708
709710
Table 1: Confusion matrix between sites categorised based on floristic composition via hierarchical clustering (rows) and sites categorised using climate and a classification tree approach (columns). The diagonal gives the number of sites that are correctly classified by climate, while the off-diagonal elements give mis-classifications (18.6%). Only non-floristically transitional sites were considered. Accuracy: 81%; Average precision: 81%; Average recall: 80%.
Amazon Forest Atlantic Forest Cerrado
Chaco
SDTF
Amazon Forest 989 6 45 0 0Atlantic Forest 3 1290 199 5 50Cerrado 58 167 357 0 50Chaco 0 7 0 76 1SDTF 0 51 65 1 408
Table 2: Confusion matrix between sites categorised based on floristic composition via hierarchical clustering (rows) and sites categorised using climate + soil and a classification tree approach (columns). The diagonal gives the number of sites that are correctly classified by climate, while the off-diagonal elements give mis-classifications (15.2%). Accuracy: 84%; Average precision: 84%; Average recall: 83%.
Amazon Forest Atlantic Forest Cerrado
Chaco
SDTF
Amazon Forest 1001 4 37 0 0Atlantic Forest 4 1331 161 4 49Cerrado 48 121 423 0 40Chaco 0 7 0 76 1SDTF 0 55 52 1 417
711
712
713
714
715
716
717
718
719
Table 3: The mean variable importance value ( one standard error) for all climatic variables included in the Random Forest analysis across 100 runs of the Bremnans’ algorithm utilizing rarefactions of the main dataset (116 sites per biome).
Environmental Variables Climate Climate + SoilMean ± SE Mean ± SE
Mean Annual Precipitation (mm) 356.81 ± 1.09 318.8 ± 1.18Temperature Seasonality (Co) 319.73 ± 1.23 287.14 ± 1.13Maximum Climatological Water Deficit (mm/yr) 273.2 ± 0.69 232.07 ± 0.71Isothermality (%) 233.29 ± 0.98 211.53 ± 0.87pH (KCl) * 188.98 ± 0.84Mean Temperature of Coldest Quarter (Co) 187.06 ± 0.95 163.07 ± 0.97Precipitation of Wettest Quarter (mm) 155.06 ± 0.56 120.57 ± 0.48Cation Exchange Capacity (cmol/Kg) * 119.89 ± 0.23Precipitation of Driest Quarter (mm) 148.46 ± 0.53 119.37 ± 0.51Precipitation of Driest Month (mm) 133.16 ± 0.49 109.94 ± 0.44Mean Annual Temperature (Co) 122.75 ± 0.71 96.15 ± 0.66Precipitation of Wettest Month (mm) 119.83 ± 0.42 91 ± 0.35Mean Temperature of Driest Quarter (Co) 106.46 ± 0.57 81.93 ± 0.49Amount of Sand (%) * 81.73 ± 0.17Maximum Temperature of Warmest Month (Co) 103.8 ± 0.33 81.69 ± 0.31Amount of Silt (%) * 76.89 ± 0.13Temperature Annual Range (Co) 101.51 ± 0.32 75.32 ± 0.23Precipitation Seasonality (%) 99.22 ± 0.24 74.3 ± 0.31Minimum Temperature of Coldest Month (Co) 99.21 ± 0.23 73.38 ± 0.37Precipitation of Warmest Quarter (mm) 98.61 ± 0.3 70.77 ± 0.18Precipitation of Coldest Quarter (mm) 97.11 ± 0.47 69.21 ± 0.25Temperature’s Diurnal Range (Co) 91.45 ± 0.19 68.67 ± 0.16Amount of Clay (%) * 65.97 ± 0.13Mean Temperature of Wettest Quarter (Co) 79.01 ± 0.22 61.57 ± 0.24Mean Temperature of Warmest Quarter (Co) 60.71 ± 0.12 46.52 ± 0.16
720
721
722
723
724
725
726
Figure 1: Hierarchical cluster of 4,103 sites in lowland (<1,000 m.a.s.l.) tropical South America and neighbouring subtropical areas based on tree species composition. Five principal higher-level groups can be observed, which were refer to as the Amazon Forest (blue), Atlantic Forest (green), Savanna (grey), Seasonally Dry Tropical Forest or SDTF (brown) and Chaco (black) biomes. See main text for details.
Chaco
SDTF
Savanna Atlantic Forest
Amazon Forest
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742743744745
746
747
748
749
750
751
752
753
754
755
756
Figure 2 - Map of Lowland Tropical South America with sites classified into biomes based on hierarchical cluster analysis of tree species composition: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow grey circles), Amazon Forest (blue squares), Chaco (inverted hollow black triangles). Sites that were revealed to be more similar floristically to a different biome from the one with which they originally clustered are here given the symbol of the floristically more similar biome.
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779780781782783
784
785
786
787
788
789
790
Figure 3 – Map of South America with a schematic representation of the biomes delimited via hierarchical cluster analysis in the present contribution (Amazon Forest, Atlantic Forest, Savanna, Chaco and Seasonally Dry Tropical Forests – SDTF). The map was created by applying the Thiessen polygons method on the categorised points presented in figure 2. See text for further details.
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819820821822
823
824
825
Figure 4: NeoTropTree sites which have a transitional/ambiguous floristic identity, as revealed by the silhouette analysis, and how they are distributed in geographic (a) and species compositional (b) spaces. In (a), sites are categorised according the biome to which they are floristically more similar. In (b), correctly classified sites are shown in the same colour scheme as Figure 2, whereas misclassified sites are represented in black and in the same shape as the sites of their biome based on the original clustering analysis. Symbols correspond to: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow grey circles), Amazon Forest (blue squares), Chaco (inverted hollow black triangles).
(b)(a)
826
827
828
829
830
831
832
833
834
835836837838839840841842
843
844
845
846
847
848
849
850
851
852
Figure 5: Distribution of sites with respect to precipitation regime. Mean annual precipitation values come from worldclim (Hijmans et al. 2005) and maximum climatological water deficit comes from Chave et al. (2014). Symbols correspond to: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow grey circles), Amazon Forest (blue squares), Chaco (inverted hollow black triangles). Modelled after Fig. 1 in Malhi et al. (2009), which suggested that savannas were drier than seasonal forests, contrary to the pattern here.
853
854855856857858859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
Figure 6: Distribution of sites in climatic space across the nine biomes proposed by Wittaker (1975) considering mean annual precipitation (cm) and mean annual temperature (Co). Numbers correspond to: Tropical rain forest (1), Tropical seasonal forest/savanna (2), Tropical and subtropical desert (3), Temperate rainforest (4), Temperate deciduous forest (5), Woodland/scrubland (6), Temperate grassland/dessert (7), Boreal forest (8), and Tundra (9). While symbols and colors correspond to: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow gray circles), Amazon Forest (blue squares), and Chaco (inverted hollow black triangles).
97
86
5
4
3
1
2
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890891892893894895896
897
898
899
900
901
902
903
904
905
Appendix to:
Silva de Miranda, P., L., Oliveira-Filho, A., T., Pennington, R., T., Neves, D., M., Baker, T.,
R., Dexter, K., G. (2018). Using tree species inventories to map biomes and assess their
climatic overlaps in lowland tropical South America
Appendix 1. Main Biomes of Lowland Tropical South America – Brief descriptions
Wet Forest Biomes (Amazon and Atlantic Forests)
All rain forests, moist forests, evergreen forests and most semideciduous forests fell
within two overarching groups in the cluster analysis, which we termed the Atlantic
and Amazon Forest biomes. While we have argued that a floristics approach can be
used to delimit biomes at continental scales where biogeographic factors are not the
main driver of turnover in species composition, it may be that the floristic
differentiation between the Atlantic and Amazon Forests is due in part to their
biogeographic isolation by the Dry Diagonal. However, to definitely determine whether
these forests represent distinct biomes, further comparative research is needed to
determine how they compare in terms of ecosystem function.
The Atlantic Forest biome can be further divided into three different floristic groups, a
northern group, completely tropical, encompassing all the coastal Atlantic forests
ranging from northeast Brazil south to the state of Rio de Janeiro; a second group,
largely sub-tropical, beginning at Sao Paulo’s coast and harbouring all of the forests
covering the South of Brazil, Uruguay, Southeast Paraguay and portions of Northeast
Argentina, especially the Missiones region; and a last group, also tropical, formed by
semideciduous forests further inland, scattered mostly across Brazil, but also present
as far west as Bolivia. This distribution matches the sensu-latissimo definition proposed
by (Oliveira-Filho et al., 2006) with the additional inclusion of forest patches amongst
the subtropical grasslands in the south of Brazil, Southern Paraguay, most of Uruguay
and Northeast Argentina. This region has been distinguished from the Atlantic Forest in
the past based on its overall physiognomy of forest patches in a grassy landscape,
which contrasts with contiguous forest. However, these forest patches clearly show
strong floristic continuity with the Atlantic Forest, as was also observed by Oliveira-
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
Filho et al. (2013), and likely have similar ecosystem function to the now heavily
fragmented Atlantic Forest.
The Amazon Forest biome does not show as clear subdivisions as the Atlantic Forest
biome. However, there is evident floristic differentiation between “terra firme” and
seasonally flooded forests, and these two subgroups can be further divided between
sites in the western Amazon (from Peru, Bolivia, Ecuador and the Brazilian state of
Acre) and the eastern Amazon (encompassing most of the Brazilian portion of the
Amazon Forest, including the states of Amazonas, Pará, Mato Grosso, Maranhão and
Roraima). These divisions between eastern and western Amazon and between “terra
firme” and seasonally flooded forests have been reported before in the literature (e.g.
Prance 1982; ter Steege et al., 2006).
The gallery forests within the Cerrado Domain do not cluster with the prevailing
Savanna biome in that Domain, nor do they form their own unique cluster. Instead,
they are floristically most similar to the most geographically proximal wet forest
biome, either the Atlantic or the Amazon Forest. Similarly, sites found in sandy coastal
areas of Brazil, often termed “restingas” or “matas de maré”, do not comprise a single
group in our hierarchical cluster, but cluster with the closest wet forest biome (Atlantic
or Amazon Forest).
Dry Biomes (Savanna, Seasonally Dry Tropical Forest and Chaco)
Our analyses confirm that the savannas distributed across LTSA form a single floristic
unit. There are no clear subdivisions within this Biome. Savanna is a disturbance driven
system, which may allow for the ready establishment of dispersing propagules of
dominant tree species and a homogenisation of the tree flora over large spatial scales.
Indeed, savannas in SA have been shown to possess a consistent set of dominant
oligarchic tree species (Bridgewater et al., 2004), which may be why clear subgroups
are not evident. In addition, the high disturbance in the system may prevent tree
communities from reaching an equilibrium or ‘climax’ in species composition, which
may inhibit sites from converging on similar species composition in similar
environments, which could in turn inhibit the formation of clear floristic groups.
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
Our analyses suggest that the SDTF scattered across lowland tropical South America
should be regarded as one single biome, as has been suggested by previous studies
(DRYFLOR, 2016; Pennington et al., 2000, 2009; Prado & Gibbs, 1993). As found by
Neves et al. (2015) and DRYFLOR (2016), our results suggest two main groups across
the Dry Diagonal, one comprising the various forests of the Caatinga Domain and the
other comprising SDTF patches scattered throughout the Cerrado Domain and into
regions of the Pantanal and Chiquitania. The Misiones floristic group here shows
greater floristic affinity with the Atlantic Forest than it does with other SDTF. The
Misiones forests receive more rainfall than other STDF (Neves et al. 2015) and are
semi-deciduous in nature (DRYFLOR 2016). Meanwhile, the Piedmont forests are found
to be floristically more similar to sites in the Chaco than to other SDTF. This is perhaps
not surprising given their proximity to the Chaco and that both environments receive
significant frost in the winter season (Neves et al. 2015).
The Chaco is floristically different, in terms of tree species composition, from other
sites across LTSA. While this difference has been noted in the past, particularly in
comparison with SDTF (Pennington et al., 2000; Prado & Gibbs, 1993; Spichiger et al.,
2004), it has often been attributed to the Chaco experiencing heavy frost. While many
of the sites in our Chaco biome do experience frost, a large number of sites in eastern
Bolivia, western Paraguay and south central Brazil (Mato Grosso do Sul state) do not
experience frost, and could be considered tropical in nature. We refer to these
northern Chaco sites as the ‘tropical Chaco’. It is floristically distinct from other SDTF
and may have different ecosystem function, but further research is needed to compare
ecosystem function in SDTF versus tropical and subtropical Chaco sites.
Chiquitania and Pantanal
Two regions that have always been a challenge to place in floristic or biome
classification schemes are the Chiquitania and Pantanal regions of eastern Bolivia and
southwestern Brazil. The Chiquitania region is the site of contact between savannas
(composed mostly of the savanna wetlands from the Pantanal region and the Llanos de
Moxos region in Bolivia), Amazon Forest, SDTF and the Chaco (Killeen et al., 2006;
Pennington et al., 2009). This region is composed of a mosaic of SDTF mixed with
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
savannas, overlying diverse old geological formations (Navarro, 2011), and its northern
portion grades into the Amazon Forest. Chiquitania is notable for its lack of endemic
plant species, which is attributed to its recent geological past and to its transitional
nature (Killeen et al., 2006). Our analyses show that sites within the Chiquitania’s
geographic range (Navarro, 2011) alternatively cluster together with the SDTF,
Savanna and Amazon Forest biomes, and that perhaps the region should not be
considered as a distinct vegetation entity on its own.
The floristic identity of forests and woodlands in the Pantanal also do not stand out as
distinct within a continental context, although such was proposed by Veloso et al.
(1991) and Navarro (2011). The Brazilian government also classifies it as a unique
Domain (IBGE, 2012). However, just like Chiquitania, the Pantanal is composed of sites
that belong to the Savanna and SDTF biomes as well as a wet forest biome, but in this
case the Atlantic Forest biome. The lack of endemic species in this region is also
evidence of its lack of floristic distinctness and recent geological history (Pott et al.,
2011)
References:
Bridgewater, S., Ratter, J. A., & Ribeiro, J. F. (2004). Biogeographic patterns , b-diversity
and dominance in the cerrado biome of Brazil. Biodiversity and Conservation,
13, 2295–2318.
DRYFLOR. (2016). Plant diversity patterns in neotropical dry forests and their
conservation implications. Science, 353, 1383–1387.
IBGE. (2012). Manual Técnico da Vegetação Brasileira (segunda ed). Rio de Janeiro:
Instituto de Brasileiro de Geografia e Estatística - IBGE.
Killeen, T., Chavez, E., Peña-Claros, M., Toledo, M., Arroyo, L., Caballero, J., …
Steininger, M. (2006). The Chiquitano Dry Forest, the Transition between
Humid and Dry Forest in Eastern Lowland Bolivia. In R. . T. Pennington, G. P. .
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
Lewis, & J. A. Ratter (Eds.), Neotropical Savannas and Seasonally Dry Forests
Plant Diversity, Biogeography, and Conservation (pp. 213–233). CRC Press.
Navarro, G. (2011). Classificación de la Vegetación de Bolivia. Santa Cruz de la Sierra:
Centro de Ecología Difusión Simón I. Patiño.
Neves, D. M., Dexter, K. G., Pennington, R. T., Bueno, M. L., & Oliveira-Filho, A. T.
(2015). Environmental and historical controls on floristic composition across the
South American Dry Diagonal.
Oliveira-Filho, A. T., Budke, J. C., Jarenkow, J. A., Eisenlohr, P. V., & Neves, D. R. M.
(2013). Delving into the variations in tree species composition and richness
across South American subtropical Atlantic and Pampean forests. Journal of
Plant Ecology, 8, 242–260.
Oliveira-Filho, A. T., Jarenkow, L. A., & Rodal, M. J. N. (2006). Floristic relationships of
seasonally dry forests of Eastern South America based on tree species
distribution patterns. In Neotropical Savannas and Seasonally Dry Forests: Plant
Diversity, Biogeography, and Conservation (pp. 159–192).
Pennington, R. T., Lavin, M., & Oliveira-Filho, A. (2009). Woody Plant Diversity,
Evolution, and Ecology in the Tropics: Perspectives from Seasonally Dry Tropical
Forests. Annual Review of Ecology, Evolution, and Systematics, 40, 437–457.
Pennington, R. T., Prado, D. E., & Pendry, C. A. (2000). Neotropical seasonally dry
forests and Quaternary vegetation changes. Journal of Biogeography, 27, 261–
273.
Pott, a, Oliveira, a K. M., Damasceno-Junior, G. a, & Silva, J. S. V. (2011). Plant diversity
of the Pantanal wetland. Brazilian Journal of Biology = Revista Brasleira de
Biologia, 71(1 Suppl 1), 265–273.
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
Prado, D. E., & Gibbs, P. E. (1993a). Patterns of Species Distributions in the Dry
Seasonal Forests of South America. Annals of the Missouri Botanical Garden,
80, 902–927.
Prado, D. E., & Gibbs, P. E. (1993b). Patterns of species distributions in the dry seasonal
forests of South America. Annals of the Missouri Botanical Garden, 80, 902–
927.
Prance, G. T. (1982). Biological Diversity in the Tropics. (N. York, Ed.). Columbia
University Press.
Spichiger, R., Calenge, C., & Bise, B. (2004). Geographical zonation in the Neotropics of
tree species characteristic of the Paraguay-Paraná Basin. Journal of
Biogeography, 31, 1489–1501.
ter Steege, H., Pitman, N. C. a, Phillips, O. L., Chave, J., Sabatier, D., Duque, A., …
Vásquez, R. (2006). Continental-scale patterns of canopy tree composition and
function across Amazonia. Nature, 443, 444–447.
Veloso, H. P., Rangel Filho, A. L. R., & Lima, J. C. A. (1991). Classificação da Vegetação
Brasileira Adaptada a um Sistema Universal. Rio de Janeiro: Fundação Instituto
Brasileiro de Geografia e Estatística, IBGE.
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
Appendix to:
Silva de Miranda, P., L., Oliveira-Filho, A., T., Pennington, R., T., Neves, D., M., Baker, T.,
R., Dexter, K., G. (2018). Using tree species inventories to map biomes and assess their
climatic overlaps in lowland tropical South America
Appendix 2. Indicator Species Analysis
Methods
In order to determine the biome affiliation of species, we used a modified version of
the phi coefficient of Tichy & Chytrý (2006) that leverages presence/absence data and
accounts for variation in sampling amongst groups. Specifically, we used the the rg
correlation index of De Cáceres & Legendre (2009). This varies from -1 to 1, with
positive values indicating a non-random association of a species with a group, or biome
in this case, and negative values indicating a non-random anti-association. To test if
the associations between a given species and biomes were significant, we randomized
occurrences across sites 1000 times and assessed if a species was found more or less
frequently in a biome than expected by chance, using an 0.05 alpha significance
threshold, with a multiple significance test (Šidák’s test) to avoid Type I error. Species
with a significant and positive association with any biome are henceforth referred to as
diagnostic species for that biome. Indicator species analyses were conducted using
functions in the indicspecies package for the R Statistical Software (De Cáceres &
Legendre, 2009).
Results:
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
In total, 8231 out of 10306 tree species were found to be significantly positively
associated with at least one biome (p < 0.05 after Šidák’s test). The Atlantic Forest has
2492 diagnostic species, the Amazon Forest has 4786 and the other Biomes – Savanna,
SDTF and Chaco – have 318, 459 and 177 respectively. Table S8 reports all species and
their association values (rg) for each Biome.
References:
Cáceres, M. De, & Legendre, A. N. D. P. (2009). Associations between species and groups of
sites: Indices and statistical inference. Ecology, 90, 3566–3574.
Tichy, L., & Chytrý, M. (2006). Statistical determination of diagnostic species for site groups of
unequal size. Journal of Vegetation Science, 17, 809–818.
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
Appendix to:
Silva de Miranda, P., L., Oliveira-Filho, A., T., Pennington, R., T., Neves, D., M., Baker, T.,
R., Dexter, K., G. (2018). Using tree species inventories to map biomes and assess their
climatic overlaps in lowland tropical South America
Appendix 3. Supplementary figures and tables
Table S2: Summary of results for silhouette analysis. The rows correspond to totals under the original classification, derived from the hierarchical clustering analysis, while the columns correspond to totals based on looking at the overall similarity of sites to the multidimensional centroid of each major group in the cluster. The diagonal corresponds to sites where the two approaches agree, while the off-diagonal elements correspond to sites where the two approaches disagree, which we consider to indicate sites that are transitional between the two biomes.
Amazon Forest Atlantic Forest Cerrado Chaco SDTF
Original Classification
Amazon Forest 1042 7 0 0 0 1049Atlantic Forest 46 1549 115 19 39 1768Cerrado 7 0 632 0 0 639Chaco 0 0 0 84 0 84SDTF 2 10 13 13 525 563Corrected Classification
1097 1566 760 116 564 4103
Table S3: Confusion matrix between sites categorised based on floristic composition via hierarchical clustering (rows) and sites categorised using climate + soil and a classification tree approach (columns), for all sites including ones identified as transitional via a silhouette analysis. The diagonal gives the number of sites that are correctly classified by climate + soil, while the off-diagonal elements give mis-classifications (20.7%). Accuracy: 79%; average precision: 78%; average recall rate: 76%.
Amazon Forest Atlantic Forest Cerrado Chaco SDTFAmazon Forest 1021 8 66 0 2Atlantic Forest 7 1281 209 11 58Cerrado 79 179 439 1 62Chaco 0 22 0 89 5SDTF 2 60 75 5 422
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
Table S4: Confusion matrix between sites categorised based on floristic composition via hierarchical clustering (rows) and sites categorised using climate and a classification tree approach (columns), for all non-geographically overlapping sites (those with centres >10 km apart). The diagonal gives the number of sites that are correctly classified by climate, while the off-diagonal elements give mis-classifications (20.3%). Accuracy: 79%; average precision: 79%; average recall rate: 77%.
Amazon Forest Atlantic Forest Cerrado Chaco SDTFAmazon Forest 812 6 63 0 1Atlantic Forest 6 1051 158 7 44Cerrado 66 137 361 1 56Chaco 0 14 1 78 4SDTF 2 42 64 4 353
Table S5: Confusion matrix between sites categorised based on floristic composition via hierarchical clustering (rows) and sites categorised using climate + soil and a classification tree approach (columns), for all sites including ones identified as transitional via a silhouette analysis. The diagonal gives the number of sites that are correctly classified by climate + soil, while the off-diagonal elements give mis-classifications (17.7%). Accuracy: 82%; average precision: 81%; average recall rate: 80%.
Amazon Forest Atlantic Forest Cerrado ChacoSDTF
Amazon Forest 1038 6 51 0 2Atlantic Forest 9 1317 174 10 56Cerrado 69 136 500 1 54Chaco 0 22 0 87 7SDTF 3 57 64 4 436
1132
1133
1134
1135
1136
1137
1138
Figure S1 – Silhouette plot with all 4103 NeoTropTree sites included in the cluster analysis. Positive Silhouette width values (Si) indicate that a site is indeed most similar, in terms of tree species composition, to the other sites in the biome it has been assigned to, whereas negative values indicate that a given site is compositionally more similar to one of the other biomes delimited through the cluster analysis than it is to the biome with which it clustered. The plot also presents the number of sites that compose each biome and their average silhouette width value (S i).
1139
114011411142114311441145
1146
Figure S2 – Map of South America with areas coloured according to Dinerstein et al., (2017), which combines ecoregions into biomes and is a reviewed and updated version of Olson et. al. (2001). The points on the map are the NeoTropTree tree species inventory sites classified into biomes by this study: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow gray circles), Amazon Forest (blue squares), and Chaco (inverted hollow black triangles) .
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165116611671168
Figure S3 – Map of South America with areas coloured according to the Domain system of IBGE (2012), which are also sometimes referred to as biomes. The points on the map are the Brazilian NeoTropTree tree species inventory sites classified into biomes by this study: Atlantic Forest (green triangles), Seasonally Dry Tropical Forest (brown circles), Savanna (hollow gray circles), Amazon Forest (blue squares), and Chaco (inverted hollow black triangles).
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186118711881189
Table S6: Confusion matrix between sites categorised according to Dinerstein et al. (2017) biome classification system (rows), which was adapted from Olson et al.(2001), and sites as categorised into biomes in this study (columns). The underlined numbers represents the sites that were assigned to matching categories between the two systems. The other elements are treated here as mis-classifications.
Existing Lowland Tropical South America Biomes according
to tree species composition
Dinerstein et al. (2017) - Adapted from Olson et al. (2001)Amazon Forest
Atlantic Forest Savanna Chaco SDTF Total
Tropical and subtropical moist broadleaf forests (tropical and subtropical humid) 994 994 117 0 41 2146
Tropical and subtropical grasslands, savannas and shrublands (tropical and subtropical semiarid) 57 403 544 99 90 1193
Tropical and subtropical dry broadleaf forests (tropical and subtropical semihumid) 21 88 78 1 420 608
Flooded grasslands and savannas (temperate to tropical fresh or brackish water inundated) 1 18 16 4 15 54
Temperate grasslands, savannas and shrublands (temperate semiarid) 0 10 0 12 0 22
Mangrove (subtropical and tropical salt water inundated) 24 51 5 0 0 801190
Table S7: Confusion matrix between sites categorised according to IBGE (2012) biome/phytogeographic domain classification system (rows), which was adapted from Veloso (1992), and sites as categorised into biomes in this study (columns). The underlined numbers represents the sites that were assigned to matching categories between the two systems. The other elements are treated here as mis-classifications.
Existing Lowland Tropical South America Biomes according to tree
species composition
Brazilian biomes Amazon Forest Atlantic Forest SavannaSDT
F Chaco TotalAmazon Forest 635 20 74 1 0 730Atlantic Forest 0 911 37 31 0 979Cerrado 35 287 539 114 2 977Caatinga 0 57 54 369 0 480Pampa 0 78 0 0 1 79Pantanal 2 9 21 18 1 51Continental water mass 85 21 4 10 1 121Oceanic water mass 6 89 2 2 0 99
1191
top related